tcwg-cleanup-stale-containers.sh: Add cleanup of ssh-agent processes.

Change-Id: Ia50672fbab9ade7922feb12a6199f4ae581c18cd

1 files changed, 22 insertions, 2 deletions

diff --git a/tcwg-cleanup-stale-containers.sh b/tcwg-cleanup-stale-containers.sh
index 6515e9e0..d757cc5e 100755
--- a/tcwg-cleanup-stale-containers.sh
+++ b/tcwg-cleanup-stale-containers.sh
@@ -10,7 +10,11 @@ Options:
   --cleanup-stopped-hours HOURS
 	Cleanup running/stopped containers that have been created more
 	than HOURS ago.  Setting HOURS to negative values will
-	run the cleanup in dry-run mode.
+	run the cleanup in dry-run mode.  Value "0" disables the cleanup.
+
+  --cleanup-ssh-agent-hours HOURS
+	Cleanup stale ssh-agent processes that were started more than
+	HOURS ago.  Value "0" disables the cleanup.
 
   --cleanup-images true/false
 	Whether to cleanup untagged images
@@ -18,6 +22,10 @@ Options:
   --cleanup-volumes true/false
 	Whether to cleanup dangling volumes
 
+  --max-containers N
+	Check that number of containers after cleanups doesn't exceed N.
+	Value "0" disables the check.
+
   --verbose true/false
 	Whether to run in verbose mode
 EOF
@@ -26,6 +34,7 @@ EOF
 
 cleanup_running_hours="-10"
 cleanup_stopped_hours="-240"
+cleanup_ssh_agent_hours="-48"
 cleanup_images=false
 cleanup_volumes=false
 verbose=false
@@ -35,6 +44,7 @@ while [ $# -gt 0 ]; do
     case $1 in
 	--cleanup-running-hours) cleanup_running_hours="$2"; shift ;;
 	--cleanup-stopped-hours) cleanup_stopped_hours="$2"; shift ;;
+	--cleanup-ssh-agent-hours) cleanup_ssh_agent_hours="$2"; shift ;;
 	--cleanup-images) cleanup_images="$2"; shift ;;
 	--cleanup-volumes) cleanup_volumes="$2"; shift ;;
 	--max-containers) max_containers="$2"; shift ;;
@@ -58,7 +68,9 @@ do_cleanup_containers ()
     local cleanup_containers=true
     local only_jenkins_containers=true
 
-    if [ "$hours" -lt "0" ]; then
+    if [ "$hours" -eq "0" ]; then
+	exit 0
+    elif [ "$hours" -lt "0" ]; then
 	hours="$((0-$hours))"
 	cleanup_containers=false
     fi
@@ -181,6 +193,14 @@ if [ ${#rm_images[@]} != 0 ]; then
     fi
 fi
 
+# This is a workaround for https://issues.jenkins-ci.org/browse/JENKINS-49097.
+# Matrix jobs leave stale ssh-agent processes, which accumulate into hundreds.
+# This cleanup has nothing to do with docker containers, but creating
+# a separate script/job for a one-liner doesn't seem to worth it.
+if [ "$cleanup_ssh_agent_hours" -gt "0" ]; then
+    killall --older-than ${cleanup_ssh_agent_hours}h -u $USER ssh-agent
+fi
+
 # Check if we have more containers than max_containers
 nb_containers=$($DOCKER ps -a | wc -l)